Comments on: Optimism and manycore computing http://storagemojo.com/2008/06/26/optimism-and-manycore-computing/ Data storage info & analysis Thu, 20 Nov 2008 23:41:04 +0000 http://wordpress.org/?v=2.6.3 By: Bill Todd http://storagemojo.com/2008/06/26/optimism-and-manycore-computing/#comment-197012 Bill Todd Sat, 02 Aug 2008 22:23:01 +0000 http://storagemojo.com/?p=726#comment-197012 I hadn't noticed your 'update' before (another example where some kind of organization that bounced an article toward the front of the list - or at least to the front of an ancillary list that people could check if they were interested - whenever it changed or an additional comment was made could be useful). Multi-core chips as currently implemented do not multiple 'real machines' make, unless the *hardware* guarantees isolation between the cores, their caches, and the memory they use in the manner that IBM's 'LPARs' and similar solutions in other high-end processors can. Until such support appears in commodity multi-core chips they aren't (functionally) any different from single-core chips in terms of being able to support safely separated OS instances: both require a hypervisor layer to do so. Solaris 'Zones', to the degree that I understand them, are just another OS protection/isolation mechanism that isolates compartments from each other only to the degree that the OS succeeds in this effort, rather than isolates multiple distinct OS instances from each other using an underlying hypervisor which is presumably somewhat more reliable - due to its layering and relative simplicity if nothing else - in this pursuit. - bill I hadn’t noticed your ‘update’ before (another example where some kind of organization that bounced an article toward the front of the list - or at least to the front of an ancillary list that people could check if they were interested - whenever it changed or an additional comment was made could be useful).

Multi-core chips as currently implemented do not multiple ‘real machines’ make, unless the *hardware* guarantees isolation between the cores, their caches, and the memory they use in the manner that IBM’s ‘LPARs’ and similar solutions in other high-end processors can. Until such support appears in commodity multi-core chips they aren’t (functionally) any different from single-core chips in terms of being able to support safely separated OS instances: both require a hypervisor layer to do so.

Solaris ‘Zones’, to the degree that I understand them, are just another OS protection/isolation mechanism that isolates compartments from each other only to the degree that the OS succeeds in this effort, rather than isolates multiple distinct OS instances from each other using an underlying hypervisor which is presumably somewhat more reliable - due to its layering and relative simplicity if nothing else - in this pursuit.

- bill

]]> By: ian http://storagemojo.com/2008/06/26/optimism-and-manycore-computing/#comment-196343 ian Thu, 03 Jul 2008 19:07:36 +0000 http://storagemojo.com/?p=726#comment-196343 "A “thin” virtualization layer atop a manycore OS - Windows 7? - could enable Microsoft to take back VMware’s market cap and reassert control of the entire OS stack." This is exactly what Solaris 10 is with zones. “A “thin” virtualization layer atop a manycore OS - Windows 7? - could enable Microsoft to take back VMware’s market cap and reassert control of the entire OS stack.”

This is exactly what Solaris 10 is with zones.

]]>